Abstract:The development of intelligent oceans requires exploration and an understanding of the various characteristics of the oceans. The emerging Internet of Underwater Things (IoUT) is an extension of the Internet of Things (IoT) to underwater environments, and the ability of IoUT to be combined with deep learning technologies is a powerful technology for realizing intelligent oceans. The underwater acoustic (UWA) communication network is essential to IoUT. The thermocline with drastic temperature and density variations can significantly limit the connectivity and communication performance between IoUT nodes. To more accurately capture the complexity and variability of ocean remote sensing data, we first sample and analyze ocean remote sensing datasets and provide sufficient evidence to validate the temporal redundancy properties of the data. We propose an innovative deep learning approach called Ocean-Mixer. This approach consists of three modules: an embedding module, a mixer module, and a prediction module. The embedding module first processes the location and attribute information of the ocean water and then passes it to the subsequent modules. In the mixing module, we apply a temporal decomposition strategy to eliminate redundant information and capture temporal and channel features through a self-attention mechanism and a multilayer perceptron (MLP). The prediction module ultimately discerns and integrates the temporal and channel relationships and interactions among various ocean features, ensuring precise forecasting. Numerous experiments on ocean temperature and salinity datasets show that Mixer-Ocean performs well in improving the accuracy of time series prediction. Mixer-Ocean is designed to support multi-step prediction and capture the changes in the ocean environment over a long period, thus facilitating efficient management and timely decision-making for innovative ocean-oriented applications, which has far-reaching significance for developing and conserving marine resources.

What problem does this paper attempt to address?

### What problem does this paper attempt to solve? This paper aims to address the issue of multi-step prediction of ocean remote sensing data, specifically focusing on the high-precision long-term prediction of the thermocline characteristics. Specifically, the paper proposes a deep learning method named **Ocean-Mixer** to tackle the following main challenges: 1. **Data Redundancy**: - There is significant data redundancy in ocean time series data, which can affect the accuracy of predictions. The paper mitigates the impact of redundant information by decomposing time sub-series. 2. **Spatiotemporal Feature Extraction**: - Existing prediction methods often handle only single-dimensional dependencies (such as time or space) and fail to effectively combine spatiotemporal features. Ocean-Mixer utilizes self-attention mechanisms and multi-layer perceptrons (MLP) to effectively extract short- to long-range temporal features and combine spatial dependencies. 3. **Multi-step Prediction**: - Traditional methods are usually limited to single-step predictions, while multi-step predictions can provide results over a longer time window, offering more decision support for ocean environment management. Ocean-Mixer is capable of multi-step time series prediction, capturing long-term trends. Through these technical means, Ocean-Mixer can more accurately predict changes in key parameters such as ocean temperature and salinity, providing important support for the development and protection of ocean resources.

A Deep Learning Approach for Multi-Step Prediction of Ocean Remote Sensing Data

Application of deep learning technique to the sea surface height prediction in the South China Sea

SDPNet: A Novel DeepLearning Method for Ocean Surface Current Prediction

Deep-Learning-Based Approach for Prediction of Algal Blooms

Predicting temporal and spatial 4-D ocean temperature using satellite data based on a novel deep learning model

Coastal Fisheries Resource Monitoring Through A Deep Learning-Based Underwater Video Analysis

Spatiotemporal Prediction of Monthly Coastal Upwelling Scenario in SST Fields Using Deep Learning Based Models

Applications of deep learning in physical oceanography: a comprehensive review

A Metadata-Enhanced Deep Learning Method for Sea Surface Height and Mesoscale Eddy Prediction

Short-Term Prediction of Global Sea Surface Temperature Using Deep Learning Networks

Deep Learning for Ocean Temperature Forecasting: a Survey

Multiple-step accurate prediction of wave energy: A hybrid model based on quadratic decomposition, SSA and LSTM

Prediction of long lead monthly three-dimensional ocean temperature using time series gridded Argo data and a deep learning method

Mid-Term Simultaneous Spatiotemporal Prediction of Sea Surface Height Anomaly and Sea Surface Temperature Using Satellite Data in the South China Sea

Dynamic Prediction of Full-Ocean Depth SSP by Hierarchical LSTM: An Experimental Result

Deep Learning–Based Prediction of Kuroshio Path South of Japan

A spatiotemporal deep learning model for sea surface temperature field prediction using time-series satellite data

A Spatiotemporal Deep Learning-Based Smart Discovery Approach for Marine Pollution Incidents from the Data-Driven Perspective

A Multi-Spatial Scale Ocean Sound Speed Prediction Method Based on Deep Learning

Big Data Driven Marine Environment Information Forecasting: A Time Series Prediction Network

A Multi-Channel LSTM Model for Sea Surface Temperature Prediction